People with dementia, regardless of the underlying pathology, present difficult management problems for caregivers, both families at home and staff in nursing home settings. Recognizing the need to limit physical and chemical restraints, nurse investigators have begun to focus on environmental manipulation as a clinical strategy for reducing the state of agitation which frequently accompanies dementia. Two major perspectives on what triggers amplification of agitated behaviors include a mismatch of environmental stimuli to the cognitive processing capacity, and altered circadian rhythms and disturbed sleep. The aims of this study of cognitively impaired institutionalized older adults exhibiting agitation are to: 1) test the effects of evening light therapy on agitated behaviors; 2) determine whether the effects of evening light therapy on agitated behaviors are associated with a) sleep consolidation, duration, and timing, b) circadian rhythm timing (body temperature), or c) nighttime urine melatonin and cortisol; 3) describe the relationship among sleep consolidation, duration and timing, circadian timing, and agitated behaviors prior to light therapy; and 4) determine whether person characteristics are related to the effectiveness of light therapy in reducing agitated behaviors. The study is a randomized cross-over experimental design with pre- and post-test of the effects of evening light using repeated measures. The sample will be 88 residents of 11 nursing homes. The intervention consists of exposure to light intensity of 3000 lux delivered over a 90 minute period via two incandescent lights mounted above the resident s bed. A placebo condition will be used consisting of a sham electromagnetic field box. The light or placebo condition will be administered for 14 days during the hours between 7-10 PM, during which time the light automatically ramps down to 0 lux in 90 minutes. Pre-post measures are taken four days before treatment or placebo and during the last four days of treatment/placebo. Pre-post measures consist of 1) observation of behaviors, 2) activity, measured on an activity monitor, 3) body temperature, measured on activity monitor, and 4) nighttime urine melatonin and cortisol. A two week washout period occurs between the treatment and/or placebo condition. Data analysis will consist of both parametric and non-parametric tests depending on variable distributions; repeated measures ANOVA, (Aim 1, 2); cosinor analysis to calculate amplitude, mesor and acrophase (Aim 2); non-parametric correlations (Aim 3); and regression (Aim 4).